ADMET prediction techniques can be particularly useful during the lead optimization phase to help fix any persistent ADMET liability in a lead series.
- Our scientists have extensive experience in building QSAR and Free-Wilson models, using project data or published data, to predict different ADMET properties for new compounds and support multi-parameter optimization during design. We can also apply established ADMET models from software vendors and open source applications to estimate properties and predict sites of metabolism.
- A range of machine learning and AI techniques can now be utilized to create and deploy predictive models to bench chemists based on local or global data. These models can be retrained and adapted throughout the project to ensure that the best predictive performance is delivered.
- To guide us in the design of molecules that can penetrate into the brain, we have implemented and validated our own version of the Central Nervous System Multiparameter Optimization (CNS MPO) approach originally developed by Pfizer, as well as the more reecnt ‘BBB Score’ from Gupta, et al. Additional literature models for drug-likeness and other properties are also available via the Squonk portal, e.g. Abbvie MPS and KiDS MPO.
- Other capabilities explored at Sygnature include PK and PK/PD modelling, as well as physiologically based pharmacokinetic (PBPK) modelling in conjunction with the DMPK group.